Research On Fault Diagnosis Of Aeroengine Rotor System Based On Vibration Analysis

The aero engine,which usually works in harsh environments,is prone to failure,and it has relatively high maintenance costs.The engine rotor system is more prone to friction,cracks and bearing damage.Therefore,research on aero-engine rotor system fault diagnosis is of great significance for aero-engine maintenance and the development of the aviation industry.According to the related research on fault diagnosis,this paper mainly carries out related researches such as multi-fault rotor system modeling,vibration signal fault diagnosis and fault mode identification.In terms of modeling a multi-fault rotor system,a dynamic model is established with reference to the characteristics of a real rotor system to solve the problem of a small amount of fault data,and a differential equation for the dynamics of the rotor system is obtained.Lateral cracks,rubbing of the rotating shaft,and the outer ring of the rolling bearing fall off The faults are numerically solved based on the variable step size Runge-kutta method to obtain the dynamic response under different faults,and the cell array method is used to extract sufficient vibration data under different faults.In terms of time-frequency fault diagnosis of vibration signals,considering the shortcomings of empirical mode decomposition that is easy to diverge at the endpoints and easy to distort the modes,this paper proposes an improved method based on cyclic stationary period extension and combined empirical mode decomposition with mutual information and The fault diagnosis method of independent component analysis uses rolling bearing and rotor system fault signals to verify the validity of the method.It can ideally eliminate the end effect and separate the low-order and high-order frequency components.In terms of rotor system fault identification and classification,combining the empirical modal decomposition method with the support method for the bearing and rotor system fault diagnosis research,the fault recognition rate is up to 94.735%.When dealing with multiple fault classifications,the results obtained by SVM are not good.Therefore,the CNN with two dimensional processing of vibration data is proposed.According to the basic principles of convolutional neural network,eight different vibration signals of the rotor system under are classified after two-dimensional processing,and the failure mode recognition rate reaches 95.29%.